COMPUTATIONAL CLASSIFICATION OF TIMBRE FOR RANKING OF TONE COLOR

Abstract

The article describes a new computational approach for timbre classification for ranking tone color based on weighted feature extraction enabling a meaningful quantitative granular timbral description. Using a quantitative approach for algorithm development paired with a normative survey and data-driven testing, the authors find the proposed method to be a highly viable approach for computational timbre classification. As an example, we illustrate the proposed method against results from a conducted perceptual experiment. Our approach allows for improved accuracy for timbral description using numerical representations. Our research results supplement and improve the existing practice of using semantic descriptors for timbral description and can be used as an assistive tool in digital music production practices. This study aims to introduce a computational approach for timbre classification, enabling granularity in the timbral description. Our study suggests the possibility of timbral classification by numerical representation as a novel method for a more accurate mode of description for timbre.